The integration of passive inductors on integrated circuits with active components and other passive components improves performance and reduces manufacturing costs. The most successful form of a monolithic passive inductor is a conductive spiral formed in a plane parallel to the substrate. Magnetic flux lines from this spiral inductor extend perpendicular to the substrate, and hence into the substrate. The spiral inductor has been particularly successful when formed on gallium arsenide (GaAs) substrates. GaAs-based processes are much more expensive than the less exotic silicon (Si) based processes. Accordingly, GaAs substrates are typically used only when Si substrates are not suitable for the application. Higher frequency (e.g. &gt;3 GHz) applications are particularly suited to GaAs implementations. At these higher frequencies, passive inductors having inductance values in the range of 1 nH or smaller are often sufficient. Such small valued inductors may be formed as spirals using little die area. Moreover, since GaAs is a semi-insulative material, losses are small nd high quality factors (Q's) can be obtained.
However, at lower frequencies where less exotic and much less expensive Si-based processes can be used (e.g. &lt;3 GHz), passive monolithic inductors tend to be less successful. At these lower frequencies, passive inductors having greater inductance values are required by circuit designs. These larger inductance values are difficult to achieve. Larger die area is required to achieve larger inductance values, other factors remaining unchanged. Moreover, Si has much higher conductivity than GaAs. Consequently, the coupling of flux lines into the Si substrate causes greater inductive losses and a lower Q when compared to a GaAs substrate. The lower Q can be somewhat compensated for by still larger inductance values which require even more die area to implement. Unfortunately, as the die area used to implement a given inductance value increases, the parasitic capacitance also increases. This leads to a lowering of self resonance that can make the inductor fail at even low frequencies.
Attempts have been made at forming types of monolithic passive inductors other than spiral inductors. One promising type of passive monolithic inductor is a coil inductor having a coil axis parallel to the substrate. This type of passive inductor is promising because a portion of the inductor flux lines do not naturally pass through the substrate. Consequently, losses can be reduced when Si substrates are used. While coil inductors are promising, they have nevertheless met with little success, and designers are too often forced to use active inductors or inductor components external to an integrated circuit chip.
The limited success of coil inductors has been due, at least in part, to an inability to generate sufficiently high inductance values for a given small die area and to reliance upon processing requirements that are incompatible with standard Si-based processing. High inductance values from a given small die area have been elusive for conventional passive monolithic coil inductors because conventional devices fail to adequately guide magnetic flux lines away from the substrate. This leads to losses which are typically compensated for with larger than desirable geometric structures that exhibit undesirably high parasitic capacitances. The use of magnetic materials to channel and guide flux lines helps, but conventional devices fail to provide complete magnetic circuits. As a result, the portion of a magnetic flux circuit that passes through a non-magnetic material exerts a great inductance-limiting influence on the resulting inductance.
Incompatible or otherwise unusual processing requirements are undesirable because they lead to low yield devices and to increased costs. One example of incompatible processing uses multiple magnetic material layers. The use of a single magnetic material layer is an atypical processing requirement that is accommodated only by accepting increased risk. Risks are increased because magnetic materials used in semiconductor processing tend to be considerably harder and to have vastly different coefficients of thermal expansion than more usual semiconductor materials. These risks are increased as more magnetic material layers are used, as the magnetic material layer nears an active semiconductor layer or as it becomes thicker. The use of multiple layers of such materials, placement of such materials near active layers, or the use of thick layers of such materials leads to low yields and increased costs.